Refactor benchmarker to rely on POJO's as inputSolution, which might not always come from an inputSolutionFile

Features that we would like:

• Directly pass dataset instances

  plannerBenchmark(Factory?).benchmark(problem1, problem2, problem3);
  

• Load all files from a directory

  <problemBenchmarks>
    <problemDirectory>data/cloudbalancing/unsolved/<problemDirectory>
  </problemBenchmarks>
  

• Load all problems from a JPA-QL query on the database mapped through JPA (won't be implemented as part of this issue)

  <problemBenchmarks>
    <jpaDataSourceUrl>...<jpaDataSourceUrl>
    <jpaQuery>...<jpaQuery>
  </problemBenchmarks>
  

• Pass a custom provider that just generates the stuff on the fly

   plannerBenchmark(Factory?).benchmark(myProblemProviderThatUsesAGenerator);
  

This all starts by changing the benchmarker's internals to rely on POJO's instead of inputSolutionFile (although we'll still support the later for backwards compatibility).